const stringSet Column::getColumns() const
{
/* check for name column */
if (this->ast == NULL) {
return stringSet();
}
return getColumns(this->ast);
}
void TableManager::aggregate(columnVector aggregateColumns, columnVector summaryColumns, Filter &filter)
{
std::vector<stringSet> colIntersect;
stringSet aCols, partCols;
/* get names (eXidYYY) of all aggregating columns */
for (auto col: aggregateColumns) {
for (auto name: col->getColumns()) {
aCols.insert(name);
}
}
Table *table;
ibis::partList parts;
size_t size = 0;
/* get names (eXidYYY) of all summary columns */
stringSet sCols;
for (auto col: summaryColumns) {
for (auto name: col->getColumns()) {
int begin = name.find_first_of('(') + 1;
int end = name.find_first_of(')');
std::string tmp = name.substr(begin, end-begin);
if (tmp != "*") { /* ignore column * used for flows aggregation */
sCols.insert(tmp);
}
}
}
size = this->parts.size();
/* filter out parts without summary columns */
for (size_t i = 0; i < this->parts.size(); ++i) {
stringSet partColumns;
Utils::progressBar( "Aggregating [1/2] ", " ", size, i );
for (size_t j = 0; j < this->parts[i]->columnNames().size(); j++) {
partColumns.insert(this->parts[i]->columnNames()[j]);
}
/* compute set difference */
stringSet difference;
std::set_difference(sCols.begin(), sCols.end(), partColumns.begin(),
partColumns.end(), std::inserter(difference, difference.begin()));
/* When all summary columns are in current part, difference is empty */
if (difference.empty()) {
parts.push_back(this->parts[i]);
} else {
std::cerr << "Ommiting part " << this->parts[i]->currentDataDir() << ", does not have column '" << *difference.begin() << "'" << std::endl;
}
}
size = parts.size();
/* go over all parts and build vector of intersection between part columns and aggregation columns */
/* put together the parts that have same intersection - this ensures for example that ipv4 and ipv6 are aggregate separately by default */
for (size_t i = 0; i < parts.size(); i++) {
/* put part columns to set */
for (size_t j = 0; j < parts[i]->columnNames().size(); j++) {
partCols.insert(parts[i]->columnNames()[j]);
}
/* initialize intersection result stringSet */
colIntersect.push_back(stringSet());
/* make an intersection */
std::set_intersection(partCols.begin(), partCols.end(), aCols.begin(),
aCols.end(), std::inserter(colIntersect[i], colIntersect[i].begin()));
#ifdef DEBUG
std::cerr << "Intersection has " << colIntersect[i].size() << " columns" << std::endl;
std::cerr << "Intersect columns: ";
for (stringSet::const_iterator it = colIntersect[i].begin(); it != colIntersect[i].end(); it++) {
std::cerr << *it << ", ";
}
std::cerr << std::endl;
#endif
/* cleanup for next iteration */
partCols.clear();
Utils::progressBar( "Aggregating [2/2] ", " ", size, i );
}
/* group parts with same intersection to one table */
ibis::partList pList;
size_t partsCount = parts.size();
bool used[partsCount];
int iterPos = 0;
/* initialise used array */
for (size_t i = 0; i < partsCount; i++) {
used[i] = false;
}
/* go over all parts (theirs intersections), empty intersections are ignored */
for (std::vector<stringSet>::const_iterator outerIter = colIntersect.begin(); outerIter != colIntersect.end(); outerIter++) {
/* work with current intersection only if it has not been used before */
/* empty intersections are allowed - we might want to show sum of data that do not match (or query sum(pkt)) */
if (used[iterPos]) {
iterPos++;
continue;
}
/* add current part */
used[iterPos] = true;
pList.push_back(parts.at(iterPos));
int curPos = iterPos;
/* add all parts that have same columns as current part and are not already used */
for (std::vector<stringSet>::const_iterator it = outerIter; it != colIntersect.end(); it++) {
if (used[curPos]) {
curPos++;
continue;
}
/* compute set difference */
stringSet difference;
std::set_symmetric_difference(outerIter->begin(), outerIter->end(), it->begin(),
it->end(), std::inserter(difference, difference.begin()));
/* When sets are equal, difference is empty */
if (difference.empty()) {
/* check that parts have same column types for aggregate columns and issue a warning if not */
/* go over aggregate columns and check the types */ /* TODO check that this is not needed for summary columns */
for (stringSet::const_iterator strIter = (*outerIter).begin(); strIter != (*outerIter).end(); strIter++) {
int pos1 = -1, pos2 = -1;
/* find index of the column in both parts */
for (unsigned int i = 0; i< parts.at(iterPos)->columnTypes().size(); i++) {
if (*strIter == parts.at(iterPos)->columnNames()[i]) {
pos1 = i;
break;
}
}
for (unsigned int i = 0; i< parts.at(curPos)->columnTypes().size(); i++) {
if (*strIter == parts.at(curPos)->columnNames()[i]) {
pos2 = i;
break;
}
}
/*
The columns with given name always exist, so there is no real need to check for '-1' value.
This is merely done to avoid programming errors.
*/
if (pos1 < 0 || pos2 < 0) {
std::cerr << "Error: an unexpected error occurred while verifying data types!" << std::endl;
break;
}
/* test that data types are same */
if (parts.at(iterPos)->columnTypes()[pos1] != parts.at(curPos)->columnTypes()[pos2]) {
std::cerr << "Warning: column '" << *strIter << "' has different data types in different parts! ("
<< parts.at(iterPos)->name() << ", " << parts.at(curPos)->name() << ")" << std::endl;
}
}
/* add table to list */
pList.push_back(parts.at(curPos));
used[curPos] = true;
}
/* don't forget to increment the counter */
curPos++;
}
#ifdef DEBUG
std::cerr << "Creating table from " << pList.size() << " part(s)" << std::endl;
std::cerr << "[" << iterPos << "]Aggregate columns: ";
for (stringSet::const_iterator it = outerIter->begin(); it != outerIter->end(); it++) {
std::cerr << *it << ", ";
}
std::cerr << std::endl;
#endif
/* create table for each partList */
if (!outerIter->empty() || aggregateColumns.empty()) {
table = new Table(pList);
columnVector aggCols;
for (auto col: aggregateColumns) {
bool isThere = true;
for (auto name: col->getColumns()) {
if (outerIter->find(name) == outerIter->end()) {
isThere = false;
break;
}
}
if (isThere) {
aggCols.push_back(col);
}
}
/* aggregate the table, use only present aggregation columns */
table->aggregateWithFunctions(aggCols, summaryColumns, filter);
table->orderBy(this->orderColumns, this->orderAsc);
this->tables.push_back(table);
}
/* and clear the part list */
pList.clear();
iterPos++;
}
}
/**
* Добавляет узел в граф и строит связааные с ним зависимости(дуги)
* Исходим из предположения, что в акутальных данных содержаться
* детерминированныее данные, в том числе и скаляры-массивы.
*/
void Graph::addNode(const Node &_arg_)
{
//cout<<"node added"<<endl;
stringSet local; // local edge vars
stringSet other; // other vars to arg
Node arg = _arg_; // will be changed
arg.setActual(createActual(arg. getIn()), createActual(arg. getOut()));
// check for return in statement
bool containReturn = false;
/*
{
cout<<"Node #"<<arg. getNumber()<<" statement: ";
arg.getAction().print(cout, "");
cout<<"Node #"<<arg. getNumber()<<" (in):"<<arg.getActualIn()<<endl;
cout<<"Node #"<<arg. getNumber()<<" (out):"<<arg.getActualOut()<<endl;
} */
if (checkStatement(arg.getAction(), false, false) == -2)
containReturn = true;
/*
* Если в предложении есть возврат, то это зависимость по управлению
* и необходимо, что бы все действия были завершены до действия возврата
* это позволит сохранить семантику исполнения программы.
*/
nodes_num++;
// first phase
stringSet foreverOut = arg. getActualOut(); // arg's out
stringSet readOut = arg. getActualOut(); // current read for consistency
stringSet readIn = arg. getActualIn(); // current read for in
TNodeSet::iterator beg = nodes. begin();
TNodeSet::iterator end = nodes. end();
unsigned long layer = init_layer - 1;
bool findDep = false;
while (beg != end)
{
Node cur = (*--end);
// forward dep
if (isForwardDepend(readIn, cur. getActualOut()))
{
findDep = true;
// form local
createForwardTransition(cur. getActualOut(), readIn, local, other);
//cout<<"Forward :"<<cur. getNumber()<<"->"<<arg. getNumber()<<":"<<local<<endl;
// cur. actual(), readIn() - in
// local, other, actual - out
readIn = other;
other. clear(); // для использования в ЛЮБОЙ следующий раз
}
else
{
// backward dep
if (isBackwardDepend(foreverOut, cur. getActualIn()))
{
//cout<<"Backward :"<<cur. getNumber()<<"->"<<arg. getNumber()<<endl;
findDep = true;
}
}
// out dep
if (isOutwardDepend(foreverOut, cur. getActualOut()))
{
findDep = true;
if (arg.getNumber() == 3)
{
//cout<<"Extended info"<<endl;
//cout<<"Forever out : "<<foreverOut<<endl;
//cout<<"cur out Node #"<<cur. getNumber()<<": "<<cur.getActualOut()<<endl;
//cout<<"readOut: "<<readOut<<endl;
//cout<<"Local: "<<local<<endl;
//cout<<"Node #"<<arg. getNumber()<<" (in):"<<arg.getActualIn()<<endl;
//cout<<"Node #"<<arg. getNumber()<<" (out):"<<arg.getActualOut()<<endl;
}
createOutwardTransition(cur. getActualOut(), readOut, local, other);
//cout<<"Outward :"<<cur. getNumber()<<"->"<<arg. getNumber()<<":"<<local<<endl;
// cur. actual(), arg. getIn() - in
// local, other, actual - inout
readOut = other;
other. clear();
}
if (findDep)
{
// create edge
Edge edge(arg. getNumber(), cur. getNumber(), packArrays(local));
//edge. print(cout); // cout
if (!local. empty())
;
//cout<<"Transit :"<<local<<endl;
addEdge(edge);
local. clear();
arg.addIn(edge. getNumber());
(*end). addOut(edge. getNumber());
if (cur. getLayer() > layer)
layer = cur. getLayer();
findDep = false;
}
else
{
if (containReturn)
{
// create empty edge
Edge edge(arg. getNumber(), cur. getNumber(), stringSet());
addEdge(edge);
arg.addIn(edge. getNumber());
(*end). addOut(edge. getNumber());
if (cur. getLayer() > layer)
layer = cur. getLayer();
} // else - do nothing
}
}
// insert node
arg. setLayer(layer + 1);
nodes. push_back(arg);
}
